Question for Tim Wessman



#8

Tim,

What algorithm does the HP-50G use to find the roots of polynomials? Is it the Jenkins-Traub algorithm? The HP-50G seem to have a polynomial roots solver that works pretty well.

Namir


#9

I assume you are talking about the PROOT command, or the Solve Poly interface? This is what is found in the source there:

** The algorithm is Laguerre iteration with stepsize control

** and is based upon the ZERPOL program contained in the thesis

** entitled "A Zero Finding Algorithm Using Laguerre's Method"

** by Brian T. Smith (directed by W. Kahan).

** The implementation is a modified version of the HP-71 Math

** ROM PROOT runtime execution code implemented by Laurence Grodd.

TW


Edited: 31 Aug 2010, 4:06 p.m.


#10

More information about the HP-71B Math ROM Polynomial Root Finder can be found in the July 1984 issue of Hewlett-Packard Journal, pages 33 through 36:

http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1984-07.pdf


#11

Thanks Gerson!

#12

Thanks Tim ... gives me a good insight of how the 50G (and other graphing machines) handle getting the roots of polynomials.

Namir

#13

Tim and Gerson,

I was looking at the HP-71B Math Pac manual and it does mention the Laguerre method as being used for that ROM (and I am sure for many other HP calculators). I have a Matlab function that implements of the Laguerre method and it works very well ... so the method is good, even though my implementation is bare bone -- not scaling or other programming tricks.

Namir

Edited: 31 Aug 2010, 10:33 p.m.


#14

Hi Namir,

You may be interested by the discussion/comparison of several Polynomial Rootfinder programs as well as the matlab sample program (Appendix A) in this document: Iterative Methods for Roots of Polynomials

Regarding PROOT on HP calculators, the following post from Bill Wickes on comp.sys.hp48 back in 1992 shows that it was quite easy to port HP 71B assembler functions to the HP-48:
HP 48 Polynomial Rootfinder

Given this, I'm a bit surprised that the results for the second test on the HP 71B (+ Math Pac) and the HP 48GX (same CPU, same algorithm, same assembler code?) or the 49G+ in the following post on this forum are somewhat different (The Turtle (HP-71B) and the Hare (HP49G+) [LONG]):

0.999999999944, 1.312E-12 for the HP 71B
0.999999994032, 2.066E-12 for the 48GX
0.999999994031, 1.441E-12 for the 49G+

... the HP 71B while being the oldest machine is the closest to the exact result (1,0)!


Edited: 5 Sept 2010, 6:03 a.m.


Possibly Related Threads...
Thread Author Replies Views Last Post
  Thanks Tim for blue HP50g's keyboard colors Carey 4 199 11-07-2011, 02:33 PM
Last Post: Jan
  Tim, Help! 15C LE emulator activation not working M. Joury 16 525 09-12-2011, 11:42 PM
Last Post: M. Joury
  OT ->Tim Wessman: TI-Nspire current draw Joerg Woerner 1 128 05-08-2011, 12:11 PM
Last Post: Tim Wessman
  Tim Wessman e-mail Hugh Evans 21 737 02-25-2006, 04:01 AM
Last Post: Marcus von Cube, Germany

Forum Jump: